Building the Intent Landscape of Real-World Conversational Corpora with Extractive Question-Answering Transformers

08/26/2022
by   Jean-Philippe Corbeil, et al.
3

For companies with customer service, mapping intents inside their conversational data is crucial in building applications based on natural language understanding (NLU). Nevertheless, there is no established automated technique to gather the intents from noisy online chats or voice transcripts. Simple clustering approaches are not suited to intent-sparse dialogues. To solve this intent-landscape task, we propose an unsupervised pipeline that extracts the intents and the taxonomy of intents from real-world dialogues. Our pipeline mines intent-span candidates with an extractive Question-Answering Electra model and leverages sentence embeddings to apply a low-level density clustering followed by a top-level hierarchical clustering. Our results demonstrate the generalization ability of an ELECTRA large model fine-tuned on the SQuAD2 dataset to understand dialogues. With the right prompting question, this model achieves a rate of linguistic validation on intent spans beyond 85 We furthermore reconstructed the intent schemes of five domains from the MultiDoGo dataset with an average recall of 94.3

READ FULL TEXT

Please sign up or login with your details

Forgot password? Click here to reset